Technical plumbing

How to refactor an old website for AEO

By Arnav Mukherjee, founder of TofuBofu · July 6, 2026

Here is a mistake I made on our own site, and I am not a beginner at this. Our early homepage led with the industries we serve, IT firms, MSPs, cybersecurity, consulting. It read well to a human. But when I ran an automated read of the page, it classified us as an IT services company, not an AI visibility platform. We had buried our own category under our audience. If a tool could not tell what we were, neither could the engines.

The fix was not a rebuild. It was a wording change and a bit of structure. That is the theme of refactoring an old site for AEO: most of what holds legacy sites back is fixable in place, in the right order, without burning the whole thing down.

51%
of B2B buyers now start their research on an AI chatbot, so an old site those engines cannot read is losing at the very first step. G2, 2026

Refactor, do not rebuild

The instinct when a site underperforms in AI search is to assume it is too old and needs replacing. Rebuilds are expensive, risky, and usually solve the wrong problem. An old site typically has three fixable issues: AI crawlers cannot read it, they cannot tell what it is, or its content does not answer the questions buyers ask. You address those in sequence, and you keep the site.

The refactor order of operations

1 Audit robots.txt, view-source, scan 2 Fix rendering & access so crawlers can read you 3 Clarify your category say what you are, plainly 4 Add structure & schema headings, FAQ, JSON-LD 5 Rewrite for buyer questions specific, answer-first pages

Step 1: Audit before you touch anything

Three quick checks tell you where the problem actually is. Open yoursite.com/robots.txt and confirm you are not blocking AI crawlers like GPTBot, ClaudeBot, PerplexityBot, or Google-Extended. View the page source on your top pages and confirm your real text is in it, not just in the rendered view. And run a scan to see what AI engines say about you today.

This diagnosis matters because the fixes are different. A crawler block or empty source is a technical problem. A confident but wrong answer about what you do is a clarity problem. A generic answer that names competitors is a content problem. Fix the one you actually have.

Step 2: Fix access and rendering first

Nothing else works until crawlers can read you. If robots.txt is blocking AI bots, unblock them. If the view-source test showed an empty shell, your content is being drawn by JavaScript in the browser and AI crawlers cannot see it, so getting your platform to serve server-rendered HTML is the priority. These two fixes can restore visibility on their own, because they take you from invisible to readable.

Step 3: Make your category unmistakable

This is the one that bit us. Old sites love a clever tagline or an audience-first headline, and both hide your category. State plainly what you are, high on the homepage and in your Organization schema, then describe who you serve. If an AI engine cannot classify you, it cannot recommend you for the right queries, no matter how good the rest of the site is. This is a wording change with outsized impact.

Step 4: Add structure and schema

Legacy pages are often a slab of prose with a hero image. Break the key ones into real headings that mirror buyer questions, clear lists, and an FAQ section marked up with schema through your platform's plugin or custom-code field. This is what lets an engine lift a clean, quotable passage from your page. It is content and settings work, not engineering.

Step 5: Rewrite for the questions buyers ask

Now the compounding work. Prune the thin, generic pages that exist only to catch a keyword, because AI answers those directly and they dilute your site. Then take your important pages and rewrite them to answer specific, qualified buyer questions, your service for a particular industry, situation, or need, with the answer stated first.

You end up with fewer, sharper, better-structured pages, which is exactly what AI engines cite and what an old bloated site usually lacks. Because the domain already has some history, these changes often show up a bit faster than they would on a brand-new site.

Which pages to refactor first

An old site can have hundreds of pages, and trying to fix them all at once is how a refactor stalls. You do not need to. A small number of pages carry almost all the AEO value, so order the work by impact and stop worrying about the rest.

First

Decision-stage pages where a competitor is named and you are not

Run a scan, find the buying-intent queries where AI names a rival instead of you, and refactor the page that should answer each. This is the highest-intent, lowest-competition work, and it maps directly to pipeline. Start here every time.

Second

Your homepage and category clarity

If AI cannot tell what you are, nothing downstream helps. Fixing the homepage and Organization schema so your category is unmistakable is a one-time change that lifts everything else.

Third

Service and industry pages with real demand

Take the pages tied to services buyers actually search for, and rewrite them answer-first with FAQ schema. One specific, structured page per genuine offering.

Skip

Thin blog archives and generic filler

Do not spend refactor time on shallow posts that exist only to catch a keyword. Prune them or leave them. They dilute the site and AI answers their topics directly now.

A simple heuristic if you want to score pages yourself: how close is this page to a buying decision, and how specific is the question it answers? High on both means refactor now. Low on both means leave it. Work top-down through that list and you will see movement long before you have touched the whole site.

Start the refactor with a diagnosis

A free scan tells you whether your old site has a rendering, clarity, or content problem.

Get your free audit

Frequently asked questions

Do I need to rebuild my old website for AEO?

Almost never. Most old sites need a refactor, not a rebuild: confirm AI crawlers can read the pages, make your category clear, add structured data, and rewrite key pages to answer specific buyer questions. A full rebuild is expensive and rarely necessary unless the site is client-rendered in a way you cannot otherwise fix.

Where do I start when refactoring an old site for AI search?

Start with an audit. Check robots.txt is not blocking AI crawlers, view the page source to confirm your content is in the HTML, and run a scan to see what AI engines currently say about you. That tells you whether you have a rendering problem, a clarity problem, a content problem, or all three, so you fix the right thing first.

What is the highest-impact change for an old site?

Usually category clarity. Many older sites bury what they actually are under audience language or clever taglines, so AI cannot classify them and never recommends them for the right queries. Stating plainly what you are, high on the homepage and in Organization schema, is often the single biggest lift, and it is a wording change, not a rebuild.

Should I delete old pages when refactoring for AEO?

Prune thin, generic pages that exist only to catch a keyword, since AI answers those directly now and they dilute your site. Keep and strengthen pages that answer specific buyer questions. The goal is fewer, sharper, better-structured pages rather than a large archive of shallow ones.

How do I add AEO structure to old pages without a developer?

Rewrite key pages to lead with the answer, use real headings that mirror how buyers ask the question, and add an FAQ section with schema through your platform's plugin or custom-code field. These are content and settings changes on most CMS platforms, not engineering. The hard part is the writing, not the code.

How long does an AEO refactor take to show results?

Fixing a crawler block or a rendering problem can restore visibility as soon as engines re-crawl, often within weeks. Content and clarity changes compound over one to three months. Because you are improving an existing site with some history, results often come a little faster than starting a brand-new domain from zero.

Sources and further reading

Keep reading: AI visibility on WordPress · Does a CMS help AI visibility? · AI visibility for a new brand